Security: Insecure YAML Deserialization via yaml.FullLoader

[bachdev]

Problem

The application uses yaml.load(f, Loader=yaml.FullLoader) to parse a configuration file (config) whose path can be controlled by a command-line argument. While FullLoader is safer than the default yaml.load without a specified loader, it still allows the construction of arbitrary Python objects. If an attacker can provide a specially crafted malicious YAML file, this can lead to arbitrary code execution on the system where the application is run. This is a severe vulnerability as it allows an attacker to execute arbitrary code with the privileges of the running application.

Severity: critical
File: PW_FT_detection/main.py

Solution

Replace yaml.FullLoader with yaml.safe_load to restrict YAML parsing to basic Python types, preventing the deserialization of potentially malicious objects.

Changes

• PW_FT_detection/main.py (modified)
• PW_FT_classification/main.py (modified)

Testing

• Existing tests pass
• Manual review completed
• No new warnings/errors introduced

[Copilot]

Pull request overview

This PR mitigates a critical RCE risk by switching YAML configuration parsing in the fine-tuning entrypoints from yaml.load(..., Loader=yaml.FullLoader) to yaml.safe_load, ensuring only basic YAML types are deserialized.

Changes:

• Replace yaml.load(..., Loader=yaml.FullLoader) with yaml.safe_load in detection config loading.
• Replace yaml.load(..., Loader=yaml.FullLoader) with yaml.safe_load in classification config loading.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
PW_FT_detection/main.py	Loads the user-provided config via yaml.safe_load to prevent unsafe object construction.
PW_FT_classification/main.py	Loads the CLI-provided config via yaml.safe_load to prevent unsafe object construction.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[bachdev]

@microsoft-github-policy-service agree